I’m seeing a new issue with Intel AI Boost/NPU processing in PhotoLab 9.11.0.
My system:
Lenovo ThinkPad X1 Carbon Gen 13
Intel Core Ultra 7 265U
Windows 11 25H2 — OS Build 26200.9106
Intel AI Boost driver 32.0.100.4841
PhotoLab 9.11.0 build 758
There have been several updates to Windows 11, the Intel NPU driver and the Intel/OpenVINO NPU stack over the last few days.
What I’m seeing
The NPU is correctly detected in Windows Device Manager and there are no device errors.
Interestingly, AI denoising appears to work correctly using Intel AI Boost.
The problem occurs when I export the processed image to JPEG. PhotoLab then fails with:
[NPU_VCL] Unrecognized device ID! 0x0
followed by errors related to parsing/updating the configuration and preparing the I/O information.
If I switch PhotoLab to use the Intel GPU instead of Intel AI Boost, the same image exports to JPEG successfully.
So at the moment the issue appears to be specifically related to the NPU processing path during JPEG export, rather than the NPU being completely unavailable.
Has anyone else seen this with the latest Intel NPU driver 32.0.100.4841 and PhotoLab 9.11?
I’m also interested to know whether the recent Intel/OpenVINO update could be related.
I have already opened a support ticket with DxO and can provide logs/details if useful.
I also checked the installed Windows App Runtime packages. The system has Microsoft.WindowsAppRuntime.2 version 2.4.0.0, as well as WindowsAppRuntime 1.8 (8000.946.1701.0). This matches the recently released Windows App SDK 2.4.0 mentioned in the forum.
The interesting part is that AI denoising appears to work with Intel AI Boost; the failure occurs specifically when exporting the processed image to JPEG. Switching to the Intel GPU allows the JPEG export to complete successfully.
Good Morning Rik - I’ve also experienced problems with PL9.11.0 after upgrading to the latest Intel drivers - error received when opening any raw files - ok with tiff and jpeg with AI Boost selected in preferences - no problems with exporting jpegs or Tiffs and opens Raw files ok, but slow, when GPU selected.
Upgraded to gfx_win_101.8974 when errors started
On holiday at the moment so reinstalled previous driver gfx_win_101.8864 with a clean install and everthing working ok again
NPU Driver is 32.0.100.4841 as with yours.
I will leave with gfx_win_101.8864 for the moment and see what the next update from Intel brings - I have had problems with Intel updates in the past that have been corrected in subsequent issues.
If you look in Intel Driver & Support app, you are able to download older drivers by clicking on “Learn more about this update”
Finally, I only picked up your post this morning by accident by having a look in my unread posts on the forum - you have posted in “windows topics/bug” rather than the main “DxO PL” catergory - you may be better moving or re-posting to main catergory to hit more forum members?
@RickBlom@dh57 , hello, I’ve installed NPU, it looks fine in device manager. However, when working with DxO PL NPU usage is always 0%. Do I miss a setting in DxO? Thanks HG
I did a lot of work with the DxO Tech Team after DxO had a few problems with Intel chips on the release of PL9, but after working with them for quite a while it must be said, after submitting a ticket, PL now works fine and as it should do after the release of PL9.6.
Since the release of PL9.6 and up to PL 9.11.1.0 I haven’t had a single error or crash on export , rendering or AI masking! …. up until the latest Intel GPU drivers as I was disscusing with RickBlom previously.
DxO Tech section told me to “select “Intel AI Boost” rather than “GPU” so that the NPU attached to the Intel processor will automatically select the correct way to route the AI instructions between the NPU and the GPU. Selecting the GPU on the preferences will in effect disable the use of the NPU and the AI Boost.”
DxO tech section also told me to expect first uses of full screen preview, loupe and exports to take longer that subsequent uses and have comfirmed to me that this is what they expect to see as the NPU pre-allocates tasks efficiently between the AI boost on the NPU and the iGPU.
Sorry I can’t really help with your comment as I haven’t really looked at NPU usage - it just works for me - selecting AI Boost gives me better performance that selectiny my iGPU
Thanks Dave, that’s very useful. Especially the information from DxO Tech that AI Boost is intended to automatically route AI workloads between the NPU and iGPU.
My situation is slightly different: with my current Intel graphics driver 101.8860 and NPU 32.0.100.4841, AI denoising appears to work with AI Boost, but JPEG export fails with [NPU_VCL] Unrecognized device ID! 0x0. Switching to GPU makes the export work.
I may try 101.8864 as a controlled test, keeping the NPU driver at 32.0.100.4841. Interesting that your rollback to 101.8864 fixed your issues.
Hi again HGF - just has a quick look at NPU with task manager when opening up PL 9.11.0 - just out of interest.
no activity on loading PL - and just a few occaisional spikes once loaded- AI masking didn’t seem to result in any loading on the NPU - only real use was up to 97% when exporting XD3 Raw image to jpeg.
maybe check you have the latest NPU driver from Intel and check your preferences in PL again?
other than that maybe submit a ticket with DxO to see if this is expected activity for your setup?
